Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.compose.market/llms.txt

Use this file to discover all available pages before exploring further.

OpenCode can load organization config from: 
GET https://api.compose.market/.well-known/opencode
The config defines one provider, compose-market, backed by @ai-sdk/openai-compatible. The remote config supplies the provider and model map. Credentials stay in OpenCode’s local /connect flow.

​
Provider config shape

The remote config uses this shape: 
{
  "config": {
    "$schema": "https://opencode.ai/config.json",
    "provider": {
      "compose-market": {
        "name": "Compose.Market",
        "npm": "@ai-sdk/openai-compatible",
        "options": {
          "baseURL": "https://api.compose.market/external/v1",
          "includeUsage": true,
          "timeout": false,
          "chunkTimeout": 120000
        }
      }
    }
  }
}
The public endpoint does not include an apiKey field. Use OpenCode’s /connect flow for the compose-market provider so OpenCode stores the credential locally, outside the remote config.

​
Models

OpenCode model keys are public catalog IDs. Examples: 
gpt-5.5
openai/gpt-5.5
@cf/meta/llama-3.3-70b-instruct-fp8-fast
OpenCode accepts a smaller model metadata schema than /external/v1/models. The remote config only includes fields OpenCode can validate; the full external catalog remains available at /external/v1/models.

​
Debug

If OpenCode starts but a model call fails, check the response headers: 
x-compose-public-model
x-compose-upstream-provider
x-compose-upstream-model
They show which catalog row the request reached.